| 1 |
/* |
| 2 |
* // Copyright (c) Radzivon Bartoshyk 4/2025. All rights reserved. |
| 3 |
* // |
| 4 |
* // Redistribution and use in source and binary forms, with or without modification, |
| 5 |
* // are permitted provided that the following conditions are met: |
| 6 |
* // |
| 7 |
* // 1. Redistributions of source code must retain the above copyright notice, this |
| 8 |
* // list of conditions and the following disclaimer. |
| 9 |
* // |
| 10 |
* // 2. Redistributions in binary form must reproduce the above copyright notice, |
| 11 |
* // this list of conditions and the following disclaimer in the documentation |
| 12 |
* // and/or other materials provided with the distribution. |
| 13 |
* // |
| 14 |
* // 3. Neither the name of the copyright holder nor the names of its |
| 15 |
* // contributors may be used to endorse or promote products derived from |
| 16 |
* // this software without specific prior written permission. |
| 17 |
* // |
| 18 |
* // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 19 |
* // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 20 |
* // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
| 21 |
* // DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE |
| 22 |
* // FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 23 |
* // DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| 24 |
* // SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| 25 |
* // CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 26 |
* // OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 27 |
* // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 28 |
*/ |
| 29 |
use num_traits::MulAdd; |
| 30 |
use std::ops::{Add, Mul}; |
| 31 |
|
| 32 |
#[inline(always)] |
| 33 |
pub(crate) fn is_integerf(x: f32) -> bool { |
| 34 |
#[cfg(any( |
| 35 |
all( |
| 36 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 37 |
target_feature = "sse4.1" |
| 38 |
), |
| 39 |
target_arch = "aarch64" |
| 40 |
))] |
| 41 |
{ |
| 42 |
x.round_ties_even() == x |
| 43 |
} |
| 44 |
#[cfg(not(any( |
| 45 |
all( |
| 46 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 47 |
target_feature = "sse4.1" |
| 48 |
), |
| 49 |
target_arch = "aarch64" |
| 50 |
)))] |
| 51 |
{ |
| 52 |
let x_u = x.to_bits(); |
| 53 |
let x_e = (x_u & EXP_MASK_F32) >> 23; |
| 54 |
let lsb = (x_u | EXP_MASK_F32).trailing_zeros(); |
| 55 |
const E_BIAS: u32 = (1u32 << (8 - 1u32)) - 1u32; |
| 56 |
const UNIT_EXPONENT: u32 = E_BIAS + 23; |
| 57 |
x_e + lsb >= UNIT_EXPONENT |
| 58 |
} |
| 59 |
} |
| 60 |
|
| 61 |
#[inline(always)] |
| 62 |
pub(crate) fn is_odd_integerf(x: f32) -> bool { |
| 63 |
#[cfg(target_arch = "aarch64")] |
| 64 |
{ |
| 65 |
(x as i32 & 1) != 0 |
| 66 |
} |
| 67 |
#[cfg(not(target_arch = "aarch64"))] |
| 68 |
{ |
| 69 |
let x_u = x.to_bits(); |
| 70 |
let x_e = (x_u & EXP_MASK_F32) >> 23; |
| 71 |
let lsb = (x_u | EXP_MASK_F32).trailing_zeros(); |
| 72 |
const E_BIAS: u32 = (1u32 << (8 - 1u32)) - 1u32; |
| 73 |
|
| 74 |
const UNIT_EXPONENT: u32 = E_BIAS + 23; |
| 75 |
x_e + lsb == UNIT_EXPONENT |
| 76 |
} |
| 77 |
} |
| 78 |
|
| 79 |
#[inline(always)] |
| 80 |
pub(crate) fn is_integer(n: f64) -> bool { |
| 81 |
#[cfg(any( |
| 82 |
all( |
| 83 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 84 |
target_feature = "sse4.1" |
| 85 |
), |
| 86 |
target_arch = "aarch64" |
| 87 |
))] |
| 88 |
{ |
| 89 |
n == n.round_ties_even() |
| 90 |
} |
| 91 |
#[cfg(not(any( |
| 92 |
all( |
| 93 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 94 |
target_feature = "sse4.1" |
| 95 |
), |
| 96 |
target_arch = "aarch64" |
| 97 |
)))] |
| 98 |
{ |
| 99 |
use crate::bits::EXP_MASK; |
| 100 |
let x_u = n.to_bits(); |
| 101 |
let x_e = (x_u & EXP_MASK) >> 52; |
| 102 |
let lsb = (x_u | EXP_MASK).trailing_zeros(); |
| 103 |
const E_BIAS: u64 = (1u64 << (11 - 1u64)) - 1u64; |
| 104 |
|
| 105 |
const UNIT_EXPONENT: u64 = E_BIAS + 52; |
| 106 |
x_e + lsb as u64 >= UNIT_EXPONENT |
| 107 |
} |
| 108 |
} |
| 109 |
|
| 110 |
#[inline(always)] |
| 111 |
#[allow(unused)] |
| 112 |
pub(crate) fn is_odd_integer_fast(x: f64) -> bool { |
| 113 |
unsafe { (x.to_int_unchecked::<i64>() & 1) != 0 } |
| 114 |
} |
| 115 |
|
| 116 |
#[inline(always)] |
| 117 |
#[allow(unused)] |
| 118 |
pub(crate) fn is_odd_integerf_fast(x: f32) -> bool { |
| 119 |
unsafe { (x.to_int_unchecked::<i32>() & 1) != 0 } |
| 120 |
} |
| 121 |
|
| 122 |
#[inline(always)] |
| 123 |
pub(crate) fn is_odd_integer(x: f64) -> bool { |
| 124 |
#[cfg(any( |
| 125 |
all( |
| 126 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 127 |
target_feature = "sse4.1" |
| 128 |
), |
| 129 |
target_arch = "aarch64" |
| 130 |
))] |
| 131 |
{ |
| 132 |
(x as i64 & 1) != 0 |
| 133 |
} |
| 134 |
#[cfg(not(any( |
| 135 |
all( |
| 136 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 137 |
target_feature = "sse4.1" |
| 138 |
), |
| 139 |
target_arch = "aarch64" |
| 140 |
)))] |
| 141 |
{ |
| 142 |
use crate::bits::EXP_MASK; |
| 143 |
let x_u = x.to_bits(); |
| 144 |
let x_e = (x_u & EXP_MASK) >> 52; |
| 145 |
let lsb = (x_u | EXP_MASK).trailing_zeros(); |
| 146 |
const E_BIAS: u64 = (1u64 << (11 - 1u64)) - 1u64; |
| 147 |
|
| 148 |
const UNIT_EXPONENT: u64 = E_BIAS + 52; |
| 149 |
x_e + lsb as u64 == UNIT_EXPONENT |
| 150 |
} |
| 151 |
} |
| 152 |
|
| 153 |
#[cfg(any( |
| 154 |
all( |
| 155 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 156 |
target_feature = "fma" |
| 157 |
), |
| 158 |
target_arch = "aarch64" |
| 159 |
))] |
| 160 |
#[inline(always)] |
| 161 |
pub(crate) fn mlaf<T: Copy + Mul<T, Output = T> + Add<T, Output = T> + MulAdd<T, Output = T>>( |
| 162 |
acc: T, |
| 163 |
a: T, |
| 164 |
b: T, |
| 165 |
) -> T { |
| 166 |
MulAdd::mul_add(a, b, acc) |
| 167 |
} |
| 168 |
|
| 169 |
#[inline(always)] |
| 170 |
#[cfg(not(any( |
| 171 |
all( |
| 172 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 173 |
target_feature = "fma" |
| 174 |
), |
| 175 |
target_arch = "aarch64" |
| 176 |
)))] |
| 177 |
pub(crate) fn mlaf<T: Copy + Mul<T, Output = T> + Add<T, Output = T> + MulAdd<T, Output = T>>( |
| 178 |
acc: T, |
| 179 |
a: T, |
| 180 |
b: T, |
| 181 |
) -> T { |
| 182 |
acc + a * b |
| 183 |
} |
| 184 |
|
| 185 |
#[inline] |
| 186 |
pub(crate) const fn rintfk(x: f32) -> f32 { |
| 187 |
(if x < 0. { x - 0.5 } else { x + 0.5 }) as i32 as f32 |
| 188 |
} |
| 189 |
|
| 190 |
#[inline(always)] |
| 191 |
pub(crate) const fn fmlaf(a: f32, b: f32, c: f32) -> f32 { |
| 192 |
c + a * b |
| 193 |
} |
| 194 |
|
| 195 |
#[inline(always)] |
| 196 |
pub(crate) fn f_fmlaf(a: f32, b: f32, c: f32) -> f32 { |
| 197 |
#[cfg(any( |
| 198 |
all( |
| 199 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 200 |
target_feature = "fma" |
| 201 |
), |
| 202 |
target_arch = "aarch64" |
| 203 |
))] |
| 204 |
{ |
| 205 |
f32::mul_add(a, b, c) |
| 206 |
} |
| 207 |
#[cfg(not(any( |
| 208 |
all( |
| 209 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 210 |
target_feature = "fma" |
| 211 |
), |
| 212 |
target_arch = "aarch64" |
| 213 |
)))] |
| 214 |
{ |
| 215 |
a * b + c |
| 216 |
} |
| 217 |
} |
| 218 |
|
| 219 |
/// Optional FMA, if it is available hardware FMA will use, if not then just scalar `c + a * b` |
| 220 |
#[inline(always)] |
| 221 |
pub(crate) fn f_fmla(a: f64, b: f64, c: f64) -> f64 { |
| 222 |
#[cfg(any( |
| 223 |
all( |
| 224 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 225 |
target_feature = "fma" |
| 226 |
), |
| 227 |
target_arch = "aarch64" |
| 228 |
))] |
| 229 |
{ |
| 230 |
f64::mul_add(a, b, c) |
| 231 |
} |
| 232 |
#[cfg(not(any( |
| 233 |
all( |
| 234 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 235 |
target_feature = "fma" |
| 236 |
), |
| 237 |
target_arch = "aarch64" |
| 238 |
)))] |
| 239 |
{ |
| 240 |
a * b + c |
| 241 |
} |
| 242 |
} |
| 243 |
|
| 244 |
#[inline(always)] |
| 245 |
pub(crate) const fn fmla(a: f64, b: f64, c: f64) -> f64 { |
| 246 |
c + a * b |
| 247 |
} |
| 248 |
|
| 249 |
/// Executes mandatory FMA |
| 250 |
/// if not available will be simulated through Dekker and Veltkamp |
| 251 |
#[inline(always)] |
| 252 |
pub(crate) fn dd_fmla(a: f64, b: f64, c: f64) -> f64 { |
| 253 |
#[cfg(any( |
| 254 |
all( |
| 255 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 256 |
target_feature = "fma" |
| 257 |
), |
| 258 |
target_arch = "aarch64" |
| 259 |
))] |
| 260 |
{ |
| 261 |
f_fmla(a, b, c) |
| 262 |
} |
| 263 |
#[cfg(not(any( |
| 264 |
all( |
| 265 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 266 |
target_feature = "fma" |
| 267 |
), |
| 268 |
target_arch = "aarch64" |
| 269 |
)))] |
| 270 |
{ |
| 271 |
use crate::double_double::DoubleDouble; |
| 272 |
DoubleDouble::dd_f64_mul_add(a, b, c) |
| 273 |
} |
| 274 |
} |
| 275 |
|
| 276 |
// Executes mandatory FMA |
| 277 |
// if not available will be simulated through dyadic float 128 |
| 278 |
#[inline(always)] |
| 279 |
pub(crate) fn dyad_fmla(a: f64, b: f64, c: f64) -> f64 { |
| 280 |
#[cfg(any( |
| 281 |
all( |
| 282 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 283 |
target_feature = "fma" |
| 284 |
), |
| 285 |
target_arch = "aarch64" |
| 286 |
))] |
| 287 |
{ |
| 288 |
f_fmla(a, b, c) |
| 289 |
} |
| 290 |
#[cfg(not(any( |
| 291 |
all( |
| 292 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 293 |
target_feature = "fma" |
| 294 |
), |
| 295 |
target_arch = "aarch64" |
| 296 |
)))] |
| 297 |
{ |
| 298 |
use crate::dyadic_float::DyadicFloat128; |
| 299 |
let z = DyadicFloat128::new_from_f64(a); |
| 300 |
let k = DyadicFloat128::new_from_f64(b); |
| 301 |
let p = z * k + DyadicFloat128::new_from_f64(c); |
| 302 |
p.fast_as_f64() |
| 303 |
} |
| 304 |
} |
| 305 |
|
| 306 |
// Executes mandatory FMA |
| 307 |
// if not available will be simulated through Dekker and Veltkamp |
| 308 |
#[inline(always)] |
| 309 |
#[allow(unused)] |
| 310 |
pub(crate) fn dd_fmlaf(a: f32, b: f32, c: f32) -> f32 { |
| 311 |
#[cfg(any( |
| 312 |
all( |
| 313 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 314 |
target_feature = "fma" |
| 315 |
), |
| 316 |
target_arch = "aarch64" |
| 317 |
))] |
| 318 |
{ |
| 319 |
f_fmlaf(a, b, c) |
| 320 |
} |
| 321 |
#[cfg(not(any( |
| 322 |
all( |
| 323 |
any(target_arch = "x86", target_arch = "x86_64"), |
| 324 |
target_feature = "fma" |
| 325 |
), |
| 326 |
target_arch = "aarch64" |
| 327 |
)))] |
| 328 |
{ |
| 329 |
(a as f64 * b as f64 + c as f64) as f32 |
| 330 |
} |
| 331 |
} |
| 332 |
|
| 333 |
#[allow(dead_code)] |
| 334 |
#[inline(always)] |
| 335 |
pub(crate) fn c_mlaf<T: Copy + Mul<T, Output = T> + Add<T, Output = T> + MulAdd<T, Output = T>>( |
| 336 |
a: T, |
| 337 |
b: T, |
| 338 |
c: T, |
| 339 |
) -> T { |
| 340 |
mlaf(c, a, b) |
| 341 |
} |
| 342 |
|
| 343 |
/// Copies sign from `y` to `x` |
| 344 |
#[inline] |
| 345 |
pub const fn copysignfk(x: f32, y: f32) -> f32 { |
| 346 |
f32::from_bits((x.to_bits() & !(1 << 31)) ^ (y.to_bits() & (1 << 31))) |
| 347 |
} |
| 348 |
|
| 349 |
// #[inline] |
| 350 |
// // Founds n in ln(đĽ)=ln(đ)+đln(2) |
| 351 |
// pub(crate) const fn ilogb2kf(d: f32) -> i32 { |
| 352 |
// (((d.to_bits() as i32) >> 23) & 0xff) - 0x7f |
| 353 |
// } |
| 354 |
// |
| 355 |
// #[inline] |
| 356 |
// // Founds a in x=a+đln(2) |
| 357 |
// pub(crate) const fn ldexp3kf(d: f32, n: i32) -> f32 { |
| 358 |
// f32::from_bits(((d.to_bits() as i32) + (n << 23)) as u32) |
| 359 |
// } |
| 360 |
|
| 361 |
#[inline] |
| 362 |
pub(crate) const fn pow2if(q: i32) -> f32 { |
| 363 |
f32::from_bits((q.wrapping_add(0x7f) as u32) << 23) |
| 364 |
} |
| 365 |
|
| 366 |
/// Round towards whole integral number |
| 367 |
#[inline] |
| 368 |
pub(crate) const fn rintk(x: f64) -> f64 { |
| 369 |
(if x < 0. { x - 0.5 } else { x + 0.5 }) as i64 as f64 |
| 370 |
} |
| 371 |
|
| 372 |
/// Computes 2^n |
| 373 |
#[inline(always)] |
| 374 |
pub(crate) const fn pow2i(q: i32) -> f64 { |
| 375 |
f64::from_bits((q.wrapping_add(0x3ff) as u64) << 52) |
| 376 |
} |
| 377 |
|
| 378 |
// #[inline] |
| 379 |
// pub(crate) const fn ilogb2k(d: f64) -> i32 { |
| 380 |
// (((d.to_bits() >> 52) & 0x7ff) as i32) - 0x3ff |
| 381 |
// } |
| 382 |
// |
| 383 |
// #[inline] |
| 384 |
// pub(crate) const fn ldexp3k(d: f64, e: i32) -> f64 { |
| 385 |
// f64::from_bits(((d.to_bits() as i64) + ((e as i64) << 52)) as u64) |
| 386 |
// } |
| 387 |
|
| 388 |
/// Copies sign from `y` to `x` |
| 389 |
#[inline] |
| 390 |
pub const fn copysignk(x: f64, y: f64) -> f64 { |
| 391 |
f64::from_bits((x.to_bits() & !(1 << 63)) ^ (y.to_bits() & (1 << 63))) |
| 392 |
} |
| 393 |
|
| 394 |
#[inline] |
| 395 |
pub(crate) const fn min_normal_f64() -> f64 { |
| 396 |
let exponent_bits = 1u64 << 52; |
| 397 |
let bits = exponent_bits; |
| 398 |
|
| 399 |
f64::from_bits(bits) |
| 400 |
} |
| 401 |
|
| 402 |
#[inline] |
| 403 |
const fn mask_trailing_ones_u32(len: u32) -> u32 { |
| 404 |
if len >= 32 { |
| 405 |
u32::MAX // All ones if length is 64 or more |
| 406 |
} else { |
| 407 |
(1u32 << len).wrapping_sub(1) |
| 408 |
} |
| 409 |
} |
| 410 |
|
| 411 |
pub(crate) const EXP_MASK_F32: u32 = mask_trailing_ones_u32(8) << 23; |
| 412 |
|
| 413 |
#[inline] |
| 414 |
pub(crate) fn set_exponent_f32(x: u32, new_exp: u32) -> u32 { |
| 415 |
let encoded_mask = new_exp.wrapping_shl(23) & EXP_MASK_F32; |
| 416 |
x ^ ((x ^ encoded_mask) & EXP_MASK_F32) |
| 417 |
} |
| 418 |
|
| 419 |
#[cfg(test)] |
| 420 |
mod tests { |
| 421 |
use super::*; |
| 422 |
#[test] |
| 423 |
fn test_is_integer() { |
| 424 |
assert_eq!(is_integer(5.), true); |
| 425 |
assert_eq!(is_integer(6.), true); |
| 426 |
assert_eq!(is_integer(6.01), false); |
| 427 |
assert_eq!(is_odd_integer(5.), true); |
| 428 |
assert_eq!(is_odd_integer(6.), false); |
| 429 |
assert_eq!(is_odd_integer(6.01), false); |
| 430 |
assert_eq!(is_integerf(5.), true); |
| 431 |
assert_eq!(is_integerf(6.), true); |
| 432 |
assert_eq!(is_integerf(6.01), false); |
| 433 |
assert_eq!(is_odd_integerf(5.), true); |
| 434 |
assert_eq!(is_odd_integerf(6.), false); |
| 435 |
assert_eq!(is_odd_integerf(6.01), false); |
| 436 |
} |
| 437 |
} |